The progress in AI development is indeed as fast as some people think, says Dario Amodei, CEO of Anthropic, which develops a family of Claude large language models.
In a recent interview with the Wall Street Journal, Amodei discussed several topics, including competition to build the most powerful AI technology and AI progress.
“I am relatively confident that over the next 2-3 years we are going to see models that show up in the workplace that consumers use, that are - yes, assistant to humans - but gradually get better than us at almost everything,” he said in the interview.
Amodei said he believed that AI could surpass “all humans at almost everything” soon after he started working in this field. Back then, he thought that it could happen in the mid-2020s, though he had serious doubts. However, in the last 3-6 months, his uncertainty was greatly reduced.
The Anthropic CEO highlighted that AI surpassing humans will have both positive and negative consequences.
In the short term, which might be anywhere between 1 and 3 years, he urges to use the benefits of comparative advantage.
“Comparative advantage is remarkable. Even if a machine does 90% of your job, the other 10% becomes super leveraged. You spend all your time doing that 10%, you do 10 times more of the 10% that you did before, and you get 10 times more done because the other 90% percent is leveraged,” he said in the interview.
In a previous interview with Lex Fridman, Amodei also predicted that artificial superintelligence could be possible in 2026 or 2027.
Others, like Open AI CEO Sam Altman and the company’s co-founder Elon Musk, believe that AI could surpass humans as soon as 2025.
Meanwhile, Demis Hassabis, chief executive of Google DeepMind, and Yann LaCun, Meta’s chief scientist and the winner of the prestigious Turing Award, leaned toward a more conservative side, saying that artificial superintelligence might be with us within a decade.
Your email address will not be published. Required fields are markedmarked